Skip to content

Latest commit

 

History

History
executable file
·
268 lines (241 loc) · 9.35 KB

CHANGELOG.md

File metadata and controls

executable file
·
268 lines (241 loc) · 9.35 KB

cuSignal 0.18.0 (Date TBD)

New Features

Improvements

Bug Fixes

cuSignal 0.17.0 (Date TBD)

New Features

  • PR #241 - Add inverse_complex_cepstrum and minimum_phase to acoustics module
  • PR #270 - Add second and third order convolutions as convolve1d2o and convolve1d3o
  • PR #274 - Add nightly benchmarks to CI

Improvements

  • PR #267 - Various optimization across all functions
  • PR #271 - Increase robustness of testing
  • PR #280 - Fixing issue when reading sigmf data from parent folder
  • PR #282 - Update README to reflect current versions

Bug Fixes

  • PR #272 Fix bug in gausspulse
  • PR #275 Improve gpuCI Scripts
  • PR #281 Fix erroneous path in CI scripts
  • PR #286 Update conda install command in notebook

cuSignal 0.16.0 (21 Oct 2020)

New Features

  • PR #185 - Add function to translate PyCUDA gpuarray to CuPy ndarray
  • PR #195 - Add explicit FIR filter functionality
  • PR #199 - Added Ampere support
  • PR #208 - Remove CuPy v8 req for Kalman filter
  • PR #210 - Add signal scope example to notebooks

Improvements

  • PR #196 - Update README for cuSignal 0.15
  • PR #200 - Add if constexpr to binary reader
  • PR #202 - Performance improvement to Lombscargle
  • PR #203 - Update build process
  • PR #207 - Add CUDA 10 compatibility with polyphase channelizer
  • PR #211 - Add firfilter and channelize_poly to documentation; remove CPU only version of channelizer
  • PR #212 - Add KF to documentation build
  • PR #213 - Graceful handling of filter tap limit for polyphase channelizer
  • PR #215 - Improve doc formating for filtering operations
  • PR #219 - Add missing pytests
  • PR #222 - Improved performance for various window functions
  • PR #235 - Improve Wavelets functions performance
  • PR #236 - Improve Bsplines functions performance
  • PR #242 - Add PyTorch disclaimer to notebook
  • PR #243 - Improve Peak Finding function performance
  • PR #249 - Update README to add SDR integration instructions and improved install clarity
  • PR #250 - Update ci/local/README.md
  • PR #256 - Update to CuPy 8.0.0
  • PR #260 - Optimize waveform functions

Bug Fixes

  • PR #214 - Fix grid-stride loop bug in polyphase channelizer
  • PR #218 - Remove fatbins from source code on GH
  • PR #221 - Synchronization issue with cusignal testing
  • PR #237 - Update conda build files so fatbins are generated
  • PR #239 - Fix mode issue in peak finding module
  • PR #245 - Reduce number of default build architectures
  • PR #246 - Remove lfiltic function
  • PR #248 - Fix channelizer bugs
  • PR #254 - Use CuPy v8 FFT cache plan
  • PR #259 - Fix notebook error handling in gpuCI
  • PR #263 - Remove precompile_kernels() from io_examples
  • PR #264 - Fix Build error w/ nvidia-smi

cuSignal 0.15.0 (26 Aug 2020)

New Features

  • PR #69 - Multi-point Kalman Filter
  • PR #144 - Added AIR-T conda recipe
  • PR #119 - Added code of conduct
  • PR #122 - GPU accelerated SigMF Reader
  • PR #130 - Reorg tests and benchmarks to match #83 code refactor
  • PR #136 - Split reader and writer in IO packages and update docs
  • PR #146 - Add compatibility with Scipy 1.5.0; Default to SciPy > 1.5
  • PR #149 - Update Jetson conda to miniforge, Fix CoC, and Add SciPy Talk
  • PR #148 - Load fatbin at runtime
  • PR #69 - Multi-point Kalman Filter
  • PR #158 - Add debug flag for debugging
  • PR #161 - Basic implementation of polyphase channelizer

Improvements

  • PR #112 - Remove Numba kernels
  • PR #121 - Add docs build script
  • PR #126 - Install dependencies via meta package
  • PR #132 - Add IO API guide Jupyter notebook
  • PR #133 - Add IO module to cusignal docs
  • PR #160 - Update KF functionality
  • PR #170 - Combine tests and benchmarks
  • PR #173 - Hardcode SOS width
  • PR #181 - Added estimation notebook

Bug Fixes

  • PR #164 - Fix typos in the rest of the example code.
  • PR #162 - Fix typo in example plotting code
  • PR #116 - Fix stream usage on CuPy raw kernels
  • PR #124 - Remove cp_stream and autosync
  • PR #127 - Fix selected documentation formatting errors
  • PR #138 - Fix overflow issues in upfirdn
  • PR #139 - Fixes packaging of python package
  • PR #143 - Fix six package missing with Scipy 1.5
  • PR #152 - Fix error in detrend related to missing indexing support with cp.r_
  • PR #150 - Fix upfirdn output len for Scipy 1.5
  • PR #155 - Update CI local docker build
  • PR #153 - Fix issue with incorrect docker image being used in local build script
  • PR #157 - Add docs for Kalman Filter
  • PR #165 - Fix Kalman Filter version check
  • PR #174 - Fix bug in KF script
  • PR #175 - Update E2E Notebook for PyTorch 1.4, Fix SegFault
  • PR #179 - Fix CuPy 8.0dev CI build error

cuSignal 0.14 (03 Jun 2020)

New Features

  • PR #43 - Add pytest-benchmarks tests
  • PR #48 - Addition of decimate for FIR ftypes
  • PR #49 - Add CuPy Module for convolve2d and correlate2d
  • PR #51 - Add CuPy Module for lombscargle, along with tests/benchmarks
  • PR #62 - Add CuPy Module for 1d convolve and correlate, along with tests/benchmarks
  • PR #66 - Add CuPy Module for 2d upfirdn, along with tests/benchmarks
  • PR #73 - Local gpuCI build script
  • PR #75 - Add accelerated lfilter method.
  • PR #82 - Implement autosync to synchronize raw kernels by default
  • PR #99 - Implement sosfilt as an alternative to lfilter

Improvements

  • PR #40 - Ability to specify time/freq domain for resample.
  • PR #45 - Refactor _signaltools.py to use new Numba/CuPy framework
  • PR #50 - Update README to reorganize install instructions
  • PR #55 - Update notebooks to use timeit instead of time
  • PR #56 - Ability to precompile select Numba/CuPy kernel on import
  • PR #60 - Updated decimate function to use an existing FIR window
  • PR #61 - Fix link in README
  • PR #65 - Added deprecation warning for Numba kernels
  • PR #67 - cuSignal code refactor and documentation update
  • PR #71 - README spelling and conda install fixes
  • PR #78 - Ported lfilter to CuPy Raw Kernel (only 1D functional)
  • PR #83 - Implement code refactor
  • PR #84 - Update minimum versions of CuPy and Numba and associated conda envs
  • PR #87 - Update lfilter documentation to clarifiy single-threaded perf
  • PR #89 - Include data types in tests and benchmarks
  • PR #95 - Add .gitattributes to remove notebooks from GitHub stats
  • PR #97 - Add pytest-benchmark to conda ymls and update conda env name
  • PR #98 - Update documentation to show pytest-benchmark usage and link to API docs
  • PR #101 - Add notebook testing to CI
  • PR #103 - Update notebooks to match new code structure
  • PR #110 - Update README for 0.14 release
  • PR #113 - Add git commit to conda package

Bug Fixes

  • PR #44 - Fix issues in pytests
  • PR #52 - Mirror API change in Numba 0.49
  • PR #70 - Typo fix in docs api.rst that broke build
  • PR #93 - Remove lfilter due to poor performance in real-time applications
  • PR #96 - Move data type check to _populate_kernel_cache
  • PR #104 - Fix flake8 errors

cuSignal 0.13 (31 Mar 2020)

New Features

  • PR #6 - Added Simple WFM Demodulation, Jupyter Notebook with SoapySDR integration
  • PR #17 - Add conda recipe and gpuCI scripts
  • PR #22 - Addition of real and complex cepstrum for speech/acoustic signal processing
  • PR #25 - Raw CuPy Module for upfirdn

Improvements

  • PR #5 - Update cuSignal install directions for Windows OS.
  • PR #9 - Update cuSignal documentation and Conda install ymls to better support Jetson Devices and prune dependencies
  • PR #11 - Update cuSignal structure to match other RAPIDS projects
  • PR #20 - Updated conda environment and README file
  • PR #26 - Cache raw CuPy kernels in upfirdn
  • PR #28 - Add use_numba documentation in upfirdn and resample_poly; remove int support
  • PR #29 - Fix typos in README
  • PR #30 - Add Apache 2.0 license header to acoustics.py
  • PR #31 - Adding stream support and additional data types to upfirdn
  • PR #32 - Enable filter coefficient reuse across multiple calls to resample_poly and performance bug fixes
  • PR #34 - Implement CuPy kernels as module object and templating
  • PR #35 - Make upfirdn's kernel caching more generic; support 2D
  • PR #36 - Set default upfirdn/resample_poly behavior to use Raw CuPy CUDA kernel rather than Numba; Doc updates

Bug Fixes

  • PR #4 - Direct method convolution isn't supported in CuPy, defaulting to NumPy [Examine in future for performance]
  • PR #33 - Removes the conda test phase
  • PR #37 - Fix docs to include all cuSignal modules
  • PR #38 - Fix version in docs configuration

cuSignal 0.1 (04 Nov 2019)

New Features

Initial commit of cuSignal featuring support for:

  • Convolution
    • convolve - all methods other than direct
    • correlate - all methods other than direct
    • convolve2d
    • correlate2d
    • fftconvolve
    • choose_conv_method
  • B-splines
    • cubic
    • quadratic
    • gauss_spline
  • Filtering
    • wiener
    • lfiltic
    • hilbert
    • hilbert2
    • resample
    • resample_poly
    • upfirdn
  • Filter Design
    • bilinear_zpk
    • firwin
    • iirfilter
    • cmplx_sort
  • Waveforms
    • chirp
    • gausspulse
    • max_len_seq
    • square
  • Window Functions
    • get_window
    • barthann
    • bartlett
    • blackman
    • blackmanharris
    • bohman
    • boxcar
    • chebwin
    • cosine
    • exponential
    • flattop
    • gaussian
    • general_cosine
    • general_gaussian
    • general_hamming
    • hamming
    • hann
    • kaiser
    • nuttall
    • triang
    • tukey
  • Wavelets
    • morlet
    • ricker
    • cwt
  • Peak Finding
    • argrelmin
    • argrelmax
    • rgrelextrema
  • Spectral Analysis
    • periodogram
    • welch
    • csd
    • coherence
    • spectrogram
    • vectorstrength
    • stft
    • lombscargle
  • Extensions of Scipy Signal
    • freq_shift - Frequency shift signal